通过优化自适应激活功能的FPGAs在边缘上的高效神经网络
Yiyue Jiang1, Andrius Vaicaitis2, John Dooley2
1Department of Electrical and Computer Engineering, Northeastern University, Boston, MA 02115, USA.
Sensors (Basel, Switzerland)
|March 28, 2024
概括
带有自适应激活功能 (AAF) 的浅层神经网络在边缘设备上实现了深度神经网络 (DNN) 的准确性. 这种定制的细分线曲线神经网络 (SSCNN) 为实时无线数据处理提供了高效的FPGA实现.
科学领域:
- 边缘计算是一种边缘计算.
- 嵌入式系统 嵌入式系统
- 机器学习 硬件加速 机器学习 硬件加速
背景情况:
- 边缘设备上的深度神经网络 (DNN) 面临着计算和记忆方面的挑战.
- 浅层神经网络提供了效率,但与DNN相比,往往缺乏准确性.
- 适应激活功能 (AAF) 对于提高浅层网络性能至关重要.
研究的目的:
- 为了证明定制的自适应激活函数 (AAF) 可以使浅层神经网络实现DNN级准确性.
- 设计一个高效的FPGA实现一个细分线曲线神经网络 (SSCNN) 使用AAF.
- 为了验证拟议的SSCNN在射频功率放大器中实时数字预扭曲的性能.
主要方法:
- 开发了一种定制的细分斜线曲线神经网络 (SSCNN) 结构,具有自适应激活功能 (AAF).
- 在FPGA上实现SSCNN,并将其与实值时间延迟神经网络 (RVTDNNs),增强型RVTDNNs (ARVTDNNs) 和DNNs进行比较.
- 使用AMD/Xilinx RFSoC ZCU111进行无线电频率 (RF) 数字预扭曲应用的实验验证.
主要成果:
- 该SSCNN实现实现了与DNN相似的准确性,同时使用的硬件资源减少了40%并且没有块RAM.
- 在RFSoC ZCU111.1.上,FPGA的实施消耗了不到3%的可用资源.
- 该解决方案使时钟频率增加到221.12 MHz,从而促进了宽带宽信号传输.
结论:
- 使用AAF定制的浅层神经网络为边缘设备提供了计算效率高,节省内存的DNN替代方案.
- 拟议的SSCNN提供了一种可行的硬件加速解决方案,用于实时RF信号处理任务,如数字预扭曲.
- FPGA的实施表明了在资源有限的边缘平台上部署准确和高效的AI模型的潜力.
更多相关视频
10:44Inherent Dynamics Visualizer, an Interactive Application for Evaluating and Visualizing Outputs from a Gene Regulatory Network Inference Pipeline
Published on: December 7, 2021
2.1K
11:18Closed-loop Neuro-robotic Experiments to Test Computational Properties of Neuronal Networks
Published on: March 2, 2015
10.3K
相关概念视频
01:25Neural Circuits
1.2K
Neural circuits and neuronal pools are two of the main structures found in the nervous system. Neural circuits are networks of neurons that work together to carry out a specific task or process. They consist of interconnected neurons and glial cells, which provide structural and metabolic support.
Neuronal pools are collections of nerve cells with similar functions and interact through chemical and electrical signals. These pools include both interneurons (the central neural circuit nodes that...
Neuronal pools are collections of nerve cells with similar functions and interact through chemical and electrical signals. These pools include both interneurons (the central neural circuit nodes that...
1.2K
01:21Association Areas of the Cortex
5.3K
Association areas are regions of the cerebral cortex that do not have a specific sensory or motor function. Instead, they integrate and interpret information from various sources to enable higher cognitive processes such as memory, learning, and decision-making. Some key association areas include the following:
Prefrontal Association Area: This area is located in the frontal lobe and is involved in planning, decision-making, and moderating social behavior. It connects with primary motor areas,...
Prefrontal Association Area: This area is located in the frontal lobe and is involved in planning, decision-making, and moderating social behavior. It connects with primary motor areas,...
5.3K